Chat With an Industry Professional

A few days ago, I had a zoom chat with a real data scientist. His name was Bismayan Chakrabarti, and he works at Credit Suisse. I had a few questions planned, so the chat was similar to an interview. I asked a few questions about his job and what would I have to do to further my interest in data science. The most informative thing I asked was “What exactly does a data scientist do?”. Bismayan gave a very detailed answer to this question, and he delved into the actual process a data scientist goes through. To start off, the problem needs to be understood, Without fully understanding the problem, you cannot make good judgments, therefore understanding the problem that you want to solve is extremely important. Next, you need to determine whether the problem can and should even be solved by machine learning. Although computers are very powerful, if the problem was that your dog keeps scratching the door, data science is not helpful in this situation. In other cases, machine learning can be used, but it would be much simpler to use another method. After that, you need to confirm that the problem is even worth pursuing given the time and energy it will take to be solved. A problem that would be considered worth pursuing is organizing a store’s inventory, since data science makes that process a lot easier. Then you perform something called a Proof Of Concept. This is primarily done to big data, and it is basically an experiment that assesses the feasibility of a solution on a sample of data before it is actually implemented to all the data. This is an essential step because if you don’t experiment first, you could accidentally spoil the entire data set. The next step after the Proof of Concept is to create models. This can be done through charts, graphs, diagrams, etc. Seeing a model can help predict and make conclusions about the output. This is another step to ensure that data science is actually useful in the situation, and will actually solve the problem. An important factor to consider before starting is to check whether your data science related solution is ethical. Controls must be put in place for any unethical elements. For example, if your experiment is involved with working with the human mind, it is crucial that nothing is done to change the physiology of the mind. Finally, data science can actually be utilized, and the solution can be put into production on the big data. Obviously there is human oversight, to make sure everything goes smoothly and the computers don’t malfunction. As can be seen, the process is quite lengthy, and this is actually the ins and outs of a data scientist’s job.